spatiotemp_pseudoabs: Generate pseudo-absence record coordinates and dates

In dynamicSDM: Species Distribution and Abundance Modelling at High Spatio-Temporal Resolution

Generate pseudo-absence record coordinates and dates

Description

Function generates specified number of pseudo-absence record co-ordinates and dates either randomly or buffered in space and time.

Usage

spatiotemp_pseudoabs(
  spatial.method,
  temporal.method,
  occ.data,
  spatial.ext,
  temporal.ext,
  spatial.buffer,
  temporal.buffer,
  n.pseudoabs = 100,
  prj = "+proj=longlat +datum=WGS84"
)

Arguments

spatial.method	a character string, the spatial method for pseudo-absence generation. One of buffer or random: can be abbreviated.
temporal.method	a character string, the temporal method for pseudo-absence generation. One of buffer or random: can be abbreviated.
occ.data	optional; a data frame, with columns for occurrence record co-ordinates and dates with column names as follows; record longitude as "x", latitude as "y", year as "year", month as "month", and day as "day". Required if either temporal.method or spatial.method is buffer.
spatial.ext	the spatial extent to randomly generate pseudo-absences within. Object from which extent can be extracted of class SpatExtent, SpatRaster, an sf polygon or a numeric vector listing xmin, xmax, ymin and ymax in order. Required if spatial.method is random, and optionally used if buffer. See details.
temporal.ext	optional; a character vector, two dates in format "YYYY-MM-DD". The first represents the start of the temporal extent and the second represents the end of temporal extent to randomly generate pseudo-absences dates within. Required if temporal.method is random, and optionally used if buffer. See details.
spatial.buffer	optional; a numeric value or vector, the radius/radii in metres to generate buffered pseudo-absence coordinates within. Only required if spatial.method is buffer. See details.
temporal.buffer	optional; a numeric value or vector, the period(s) in days to generate buffered pseudo-absence dates within. Only required if temporal.method is buffer . See details.
n.pseudoabs	optional; a numeric value, the number of pseudo-absence records to generate. Default; 100.
prj	a character string, the coordinate reference system of input occ.data co-ordinates. Default is "+proj=longlat +datum=WGS84".

Details

Below we outline the various approaches to generating pseudo-absences through space and time available in the dynamicSDM package. To select the appropriate pseudo-absence generation approach and buffer size, there are many considerations. We recommend seeking the appropritae literature to inform your decision when species distribution modelling (Barbet-Massin et al., 2012, Phillips et al., 2009, Vanderwal et al., 2009).

Value

Returns data frame of pseudo-absence coordinates and dates.

Spatial buffer

If spatial.method is buffer, then the pseudo-absence record co-ordinates are randomly generated in a buffered area defined either by

• single numeric value for spatial.buffer - anywhere between the occurrence record and the circular distance surrounding this point (as specified in metres).

• two numeric values for spatial.buffer - anywhere between the closest radius from the occurrence record and the furthest away radius (as specified in metres).

For example, if spatial.buffer = c(3000,10000), then pseudo-absence co-ordinates are randomly generated at least 3000m radius away from occurrence record co-ordinate but within 10000m radius. Whereas, if spatial.buffer = 10000, then pseudo-absence co-ordinates are randomly generated anywhere between 0m and 10000m radius from the occurrence record.

If spatial.ext is given too, then the generated pseudo-absences are not only constrained to the buffered area but also to this extent. For instance, if occurrence records are coastal, you may want to clip buffers to only terrestrial regions using a country polygon given in spatial.ext.

Spatial random

If spatial.method is random, then the pseudo-absence record co-ordinates are randomly generated across spatial.ext object given.

If spatial.ext is a sf polygon or SpatRaster (mask if possible before input) then these shapes are used, instead of a simple rectangular extent (SpatExtent). Therefore, inputting one of these objects will allow for more specific pseudo-absence generation.

For example, inputting an sf polygon of a specific countries will ensure co-ordinates are terrestrial, whereas an extent (xmin, xmax, ymin, ymax) that encompasses these countries may result in the generation of pseudo-absence records in inappropriate areas, such as oceans or non-study-area countries.

Temporal buffer

If temporal.method is buffer, then pseudo-absence record dates are randomly generated between in a period defined by:

• single numeric value for temporal.buffer - any date between the occurrence record date and the total number of days specified prior or post.

• two numeric values for temporal.buffer - any date between the closest and furthers away number of days specified.

For example, if temporal.buffer = c(14,30), then pseudo-absence dates randomly generated at least 14 days from occurrence record dates but within 30 days. Whereas if temporal.buffer = 30, pseudo-absence dates are randomly generated anywhere between 0 and 30 days prior or post the occurrence record date.

If temporal.ext is given too, then the generated pseudo-absence dates are not only constrained to the buffer period but also to this temporal extent. For instance, an occurrence record recorded at the start of temporal.ext with 7 day buffer, may result in generated pseudo-absences outside of the temporal extent of the study.

Temporal random

If temporal.method is random, then pseudo-absence record dates are randomly generated within the two temporal.ext dates given.

References

Barbet-Massin, M., Jiguet, F., Albert, C. H., Thuiller, W. J. M. I. E. & Evolution 2012. Selecting Pseudo-Absences For Species Distribution Models: How, Where And How Many? 3, 327-338.

Phillips, S. J., Dudik, M., Elith, J., Graham, C. H., Lehmann, A., Leathwick, J. & Ferrier, S. 2009. Sample Selection Bias And Presence-Only Distribution Models: Implications For Background And Pseudo-Absence Data. 19, 181-197.

Vanderwal, J., Shoo, L. P., Graham, C. & Williams, S. E. 2009. Selecting Pseudo-Absence Data For Presence-Only Distribution Modeling: How Far Should You Stray From What You Know? Ecological Modelling, 220, 589-594.

Examples

data("sample_filt_data")


spatiotemp_pseudoabs(
 sample_filt_data,
 spatial.method = "random",
 temporal.method = "random",
 spatial.ext = c(20, 36, -35, -12),
 temporal.ext = c("2011-01-01", "2017-01-01")
)

dynamicSDM documentation built on June 28, 2024, 5:08 p.m.